chess engines

  

chess engines


Back to forum


Andrew Stephenson    (2007-09-10)
chess engines

Following Thibault's comments about Rybka having changed correpondence chess I purchased the rybka engine. It is a very good program but I dont see how it has affected cc anymore than Fritz. I have used Fritz 8 and 9 extensively for analysis and have until now no experience of other engines. Although I have just downloaded Toga II which is an excellent engine (and free!!) if anyone wants a free engine this is a top program that downloads in seconds and is up there with the commercial programs. I noticed Rybka seems more conservative evaluating positions than Fritz. However it has blind spots. For an example taken from the current chess cafe "Yasser Annotates" (Ivanchuk Seirawan 1990)after 1 e4 c6 2 d4 d5 3 Nc3 dxe 4 Nxe Bc5 5 Ng3 Bg6 6 h5 h6 7 Nf3 Nd7 8 h5 Bh7 9 Bd3 Bxd 10 Qxd Ngf6 11 Bf4 e6 12 0-0-0 Be7 13 c4 b5 Black offers a pawn my reaction is not to take - otb I would never take. Why open the c file for black and grabbing the pawn by Qxb5 looks risky with only 2 pawns to cover the king and open b and c files. Fritz prefers 14 c5 with 14 cxd followed by 15 Kb1 as 2nd choice after 3 minutes ply 15 depth Rybka r chooses 14 cxb cxb then 15 Qxb5?! even after 1 hour at ply 19! In cc I would look at 14cxb and 15 Qxb5 to see if I could survive and win with the extra pawn but working with Fritz it takes but a few minutes to see black has compensation after 15 Qxb5 Nd5. When 16 Be5 gives an inferior endgame for white and 16 Bd2 Rb8 gives an attack for black. The top professionals work with a range of programs Fritz, Junior, Shredder, Hiracs and Rybka to generate ideas. Does anyone have any views on these other program's characteristics?


Ilmars Cirulis    (2007-09-10 21:42:32)
...

1.e4 c6 2.d4 d5 3.Nc3 dxe4 4.Nxe4 Bf5 5.Ng3 Bg6 6.h4 h6 7.Nf3 Nd7 8.h5 Bh7 9.Bd3 Bxd3 10.Qxd3 Ngf6 11.Bf4 e6 12.O-O-O Be7 13.c4 b5


Jason Repa    (2007-09-10 22:20:52)
chess engines

Couldn't you have waited until our game is over before buying Rybka? lol! Fruit 2.3 is the strongest free UCI engine that I know of.


Thibault de Vassal    (2007-09-10 22:42:32)
chess engines

First of all, every chess engines is a choice, ie. about selectivity, and has consequences & weaknesses in particular positions. Fruit/Toga algorithm is really good but it has probably still much less chess knowledge than Rybka. I think Rybka's algorithm is really better also, Vasik Rajlich added some "human features" while other programs still think about chess like mathematics.

Rybka changed correspondence chess because Fritz or Junior (very strong chess machines) added to a good chess player makes a good centaur while Rybka is "almost" a centaur itself... Consequently a weak chess player can reach quite easily a 2100/2200 rating. That's the main reason...

Of course Rybka will always make some bad moves, but it/she builds an advantage move after move against other engines in most cases in 'calm' positions. HIARCS, that was told during a long time to think most like a human, was not a strong 'chess machine'. I don't know much about Hiarcs 11/12, but Rybka is a major improvement in this way IMO.


Jason Repa    (2007-09-10 22:48:06)
chess engines

I don't think it's a coincidence that the strongest commercial program was designed by a strong IM chess player.

It's always amazed me that the majority of chess engine programmers know very little about the game of chess itself, such as Stefan Meyer-Kahlen's Shredder. I think there comes a point in the decision making process where your human chess knowledge because an important factor.


Ilmars Cirulis    (2007-09-10 22:50:33)
Jason

Truth in your words.


Andrew Stephenson    (2007-09-10 23:31:53)
chess engines

Hi Jason As Thibault post indicates Toga is a fruit flavour ie fruit variant. Check CEGT rating list for single version engines (http://www.husvankempen.de/nunn/rating.htm) Toga II is at no3 and Fruit 2.3 is at no 5 position. As for our game no engine is ever going to save me! Maybe 28 ..Rc8 would have held but earlier on 14 exd6 Bf6 15 Bxf8 Kxf6 16 Qf3 looks really disgusting for black. I am not going to blame the opening but Fritz and I are back to the Najdorf! Apparently Aagard had a book out on offbeat sicilianl ines in which he could find no way for white to gain an advantage in this Prins line hmm Im sceptical. Thibault - thanks for expanding on your comments. I note your point about calm positions. The game I cited is the sort of wild position where Rybka is not so good later in the game it is convinced that 19 ..Nxf2 is good for black a move which loses and which fritz rejects fairly quickly.


Thibault de Vassal    (2007-09-10 23:43:36)
Rybka

Rybka is clearly not designed for wild positions. In this case, knowledge is not efficient, Junior/Fritz are !


Jason Repa    (2007-09-10 23:50:45)
chess engines

What about shredder 10 as far as wild positions go?


Thibault de Vassal    (2007-09-10 23:56:42)
Shredder 10

Shredder is really not strong enough in this case IMO. Shredder is best in calm positions where there's a good move to find many moves in advance...


Thibault de Vassal    (2007-09-11 00:03:00)
CEGT Best Versions 40/20 (AMD 4200+)

1. Rybka 2.3.2a (2975) , 2. Hiarcs 11.1 (2831) , 3. Toga II 1.3.4 (2829) , ... 6. Fritz 10 (2826)

This gap (144 points) is really extraordinay...


Andrew Stephenson    (2007-09-11 22:04:32)
Chess comp ratings

Rybka is most dominant in the single processor versions when it comes to multi processor versions the gap is less and suddenly Zap chess appears. I am not sure why this is. Another point thats interesting is Rybkas endgame evaluations there is some significant difference with Fritz here. It seems a bit overoptimistic but sometimes it seems right. Finally a striking point when using Rybka is how comparatively few positions it is analysing per second compared to Fritz maybe only 10%. This illustrates Thibaults point about Rybka's far greater reliance on chess knowledge.


Thibault de Vassal    (2007-09-11 22:12:31)
Chess comp ratings

"when it comes to multi processor versions the gap is less" .. maybe because the Rybka mp version in CEGT rating list uses 2 CPU while Zap! and other engines use 4 !?


Andrew Stephenson    (2007-09-12 07:36:30)
chess engines endgame play

Right Thibault! I am becoming more impressed with Rybka's endgame knowledge. It seems to have the extra pawn on just one side situations well understood. Is there any engine that is recognised as being the strongest at endgames? This is certainly an area where cc has helped me enormously as it has forced me to get some endgame books (and actually study them!)


Thibault de Vassal    (2007-09-12 07:36:52)
chess engines endgame play

" It seems to have the extra pawn on just one side situations well understood." .. It is undoubtly one of the keys of its success.

As far as I know, Shredder has always been said to be the strongest engine in the endgames.